feedback from manual verification in a village containing ~1000 addresses :
only one is located at the entrance, so it's unable to use it to mark entrance.
more than 90% of the localisation are located inside the true building outline.
~5% are outside the outline. However, it is often possible to find the building by geographical proximity.
~5% of the building outline are missing in osm, I haven't checked the localisation accuracy with satellite imagery.
4 localisations inside a outline are erroneous : - a swap of address between 2 building. I checked on the ground and asked a inhabitant, the sign on the ground is correct, gwr is wrong. - 2 are located on the outline of a building of the neighbour that sometimes does not have an address (for example, a garage)
a few addresses are duplicate. I have to check if it's duplicate in the original file or if it's a handling error on my part.
some correspond to buildings whose construction has not yet begun or just begun.
some localisation are located on virgin area without any construction project visible on the ground.
2 localisation are totally erroneous (one is located at 100m from the building, the other is 300m far away and in a wrong street)
I therefore think that the data is imperfect but very useful (about 1% error and 5% data to be reworked). However, the quality of the data may vary greatly from one municipality to another, since each municipality seems to have the freedom of method and persons in charge of measurements. let's wait to know what the licence will allow, even if in this village, survey for the building numbers is almost finished. but even in these cases, an integration with osmosis could easily detect new or missing address.
Le 21. 12. 17 à 10:01, Simon Poole a écrit :
I've re-projected the data and split it up in to municipalities, you can get the zipped ESRI shapefile from
http://qa.poole.ch/addresses/GWR/n.zip
by replacing n with the BfS municipality number.
Notes: a) as already said the use terms are not quite clear right now, and likely won't be before Christmas, but obviously you can familiarize yourself with the data now. Quality seems to vary a bit, in principle the nodes should be where the entrances are, but obviously in many cases that is only very rough b) what we want to do with the data in the end is unclear, and likely this will have to be decided on canton by canton c) if we decide to import the data we need to follow the import process d) we are already importing the address data from the Canton of Berne, and it would likely be helpful to finish that off before embarking on any thing new
Simon
Am 19.12.2017 um 18:19 schrieb Simon Poole:
I had contact with the BfS yesterday and clear terms of use should be forthcoming.
If I get around to it before Christmas, I'll see if I can produce a version of the data re-projected to WGS84 coordinates.
Simon
On 16. Dezember 2017 15:00:14 MEZ, Simon Poole simon@poole.ch wrote:
I was originally going to remind everybody, particularly those resident in the canton Berne, that we have an ongoing project to import addresses and trace buildings that is not complete yet http://wiki.openstreetmap.org/wiki/Canton_of_Bern_Address_Import and then I was going to point out that besides the addresses from the canton of Zürich that will become available at the beginning of January, that from our correspondence with the BfS we would be expecting the first address data from the GWR to be available for download in the next quarter too. However yesterday we were a bit surprised by the announcement that data from the GWR would become available immediately in a provisional form (this is what we had actually been actually asking the BfS to do, as there is no clear reason to wait till the "official" v! ersion is prepared by swisstopo, however they hadn't responded in a positive way). The downside is that, beside swisstopo and the BfS creating the impression of a not really well thought out, rather uncoordinated effort, it is currently not clear what the actual licence/use conditions are (there are lots of links being sent around which lead to conflicting, sometimes year old information). We'll see if we can clarify this next week. I've generated a short comparison of where we are in OSM currently vs. the GWR data:http://qa.poole.ch/addresses/ch_canton Noteworthy points: - last year roughly end of the year we had 667'807 addresses vs. 868'960 now, an increase of a good 120'000. Most of this is due to Berne increasing from 88'817 to 190'243 (actually the very small increase in the other cantons is slightly disappointing) - th! e weak cantons are those that we expected - the GWR numbers are lower than those in the available cantonal datasets mainly due to a more restrictive definition of what an address is. As soon as we've clarified the licence situation, we can have a look what the best way will be to utilize the data one way or another. Simon